1. Chain-based Reconfigurable Robots: SuperBot and it’s applications
Ilknur Kaynar-Kabul
Fall 2006

2. Overview
SuperBot
A Deployable, Multi-Functional, and Modular Self-Reconfigurable Robotic System
Distributed Control of the Center of Mass of a Modular Robot
Mark Moll, Peter Will, Maks Krivokon, and Wei-Min Shen.
In Proc. 2006 IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, Beijing, China, October 2006.
Multimode Locomotion via SuperBot Robots
Wei-Min Shen, Maks Krivokon, Harris Chiu, Jacob Everist, Michael Rubenstein, and Jagadesh Venkatesh
In Proc. 2006 IEEE Intl. Conf. on Robotics and Automation, pp. 2552–2557, Orlando, FL, 2006.

3. Self-reconfigurable robots
Lattice-based reconfigurable robots
Chain-based reconfigurable robots
Polybot
Conro
SuperBot
Hybrid systems
M-TRAN module
Tetrobot

4. SuperBot
SuperBot is a modular robot that consists of many reconfigurable modules that can demonstrate multifunction and reconfiguration [Salemi 2006]
SuperBot is being designed for NASA space exploration programs

5. SuperBot Each module has 3 revolute joints 6 genderless connectors
2 Atmega 128 CPUs
Some modules have wireless capabilities, video cameras

6. SuperBot
More flexible, mobile and efficient compared to the existing robots
A module can perform different gaits (e.g., caterpillar, sidewinder, push-and-pull, etc.) and turn and flip without any external help
Modules can be packaged in a way that is appropriate for transportation

7. Distributed Control of the Center of Mass of a Modular Robot
Mark Moll, Peter Will, Maks Krivokon, and Wei-Min Shen.
In Proc. 2006 IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, Beijing, China, October 2006.

8. Motivation
Much of work on modular and self-reconfigurable robots focuses on
Specific design of robots
Reconfiguration planning
Gait development
Few work on locomotion of modular robots in the presence of uncertainty - uneven and unknown terrain.

9. Idea of the paper
A robot can prevent itself from falling over by controlling the center of mass (COM)
Uses a gait only as a guideline for locomotion
Uses contact information & mass information to ensure a stable pose at all times.

10. Overview of the approach (1)
Presents a distributed and decentralized algorithm that computes the mass properties of the robot at each step
Modules compute the total mass, the center of mass (COM) and the inertia tensor
This information enables a module to compute joint displacements that will move the COM towards a desired position

11. Overview of the approach (2)
A gait is specifies where the COM needs to go and which leg needs to be moved, rather than specifying joint angle for every module.
Advantage:
Simplify the specification of a gait
Allow a modular robot to move over uneven terrain

12. Main issues
Computing the mass properties
Stabilizing Behavior

13. Computing the mass properties
Assumption: the modules are connected to form a tree-like structure, i.e. there are no loops
Each module computes the mass properties of the whole system
Based on its own state and on information it receives from its neighbors
It receives an estimate of the mass properties from a given connector of just the modules that are connected (directly or indirectly) to that connector

14. Computing the mass properties
A module sends new estimate to its neighbors when the modules move
If the modules do not move, the modules will eventually all converge to the true mass properties and stop sending updates to each other

15. Algorithm for Mass Computation

16. Algorithm for Mass Computation
After d iterations of the main loop, each module will have computed the correct COM, assuming the modules do not move
d: largest tree distance between 2 modules

17. Stabilizing Behavior To stabilize an arrangement of modules
Change the joint angles in the modules OR
Rearrange the modules OR
Combination of both
Option 2 can be slower than option 1

18. Stable configuration for a simple module
General idea: A configuration is stable if the contact forces can balance the gravitational force
Simple case: One point of contact and no friction
Stable if the center of mass lies on the support line
Support line: the vertical line through the point of contact
If it is not stable, then each module should adjust its joint angles

19. Simple case: Revolute joint
Consider one revolute joint: One side of the joint is connected to the contact point and the other side attached to it move along an arc of a circle

20. Simple case: Revolute joint
p1: COM of the part of the system that remains fixed
p2: COM of the part of the system that is going to be rotated
q: the position of the joint
w = p2 − q
Rθ is a 3-by-3 rotation matrix representing a rotation of θ radians about u.

21. Stabilizing all revolute joints
Finding optimal displacements for all joints simultaneously is very difficult
Solution: Use an approximate solution which tends to converge to a desired configuration very quickly.
Each joint computes its own optimal displacement independently of each other

22. Solving oscillation problem
This solution computes a desired direction to move in for all modules
Problem: Modules can oscillate around the support line due to the momentum
Solution: 2 heuristics
Based on the distance between the estimated COM and the support line
Based on momentum

23. Heuristic 1: Distance based
Reduce the gains as the COM gets closer to the support line, so that the robot does not overshoot the goal position.
Proportional gain is adjusted as follows:
c0 and c1 are constants
dsupport is the distance to the support line
KP0 is the nominal proportional gain

24. Heuristic 2: Momentum based
An ensemble of modules should not gain too much momentum
For each joint, consider the mass and the distance to the joint of the COM of the modules that will be moved by this joint
Proportional gain is adjusted as follows:

25. Simulation Results Random trees of modules are used as robots
20 modules divided into 4 branches of 5 modules
Each module has 3 DOF, the whole tree has 60 DOF
The root is always in vertical direction and fixed to the ground

26. Simulation Results
To evaluate the performance, distance between the COM and the support line as function of time is used
Tested on 3 different control schemes:
Default: The gains on all modules are identical and constant
Distance: The gains depend on the estimated distance to the support line
Momentum: The gains depend on the momentum

27. Performance for Robot (a)

28. Performance for Robot (b)

29. Performance for Robot (c)

30. Performance for Robot (d)

31. Conclusion
Presents the feasibility of using distributed control to move the COM of a modular robot to a desired position
Control methods with heuristics move the COM to a desired position
No control method outperforms the others
Momentum heuristic gives the best overall behavior
All methods exhibit the desired behavior most of the time

32. Future work
The performance can be improved if each module computes the optimal joint angles for all three joints simultaneously
Inertia tensor can be used in balancing the behavior
External forces, such as gravity and friction, at the contact points can be taken into account

33. Multimode Locomotion via SuperBot Robots
Wei-Min Shen, Maks Krivokon, Harris Chiu, Jacob Everist, Michael Rubenstein, and Jagadesh Venkatesh
In Proc. 2006 IEEE Intl. Conf. on Robotics and Automation, pp. 2552–2557, Orlando, FL, 2006.

34. Overview
Presents SuperBot for multiple locomotion modes based on reconfigurable modules
Shows the validity of the SuperBot for
the movements of forward, backward, turn, sidewinder, maneuver, and travel on batteries up to 500 meters on a flat terrain

35. Multimode locomotion
Multimode locomotion : Ability to use different moving modes in different environments.
“climb” if it is to go up a slope
“run” if it is to cover more distance with less energy
“balance” if the terrain is rugged and uneven
“get up on feet” if it fell down by mistake

36. Multimode locomotion
To support multimode locomotion, a robot must have at least four capabilities.
it must be able to perform different locomotion mode.
it must be able to recover from unexpected locomotion failures.
it must be able to shift from one mode to another.
it must be able to choose the correct mode for the correct environment.
This paper focuses

37. Multimode locomotion
2 competing and even conflicting criteria for multimode locomotion:
the robot must be general
To deal with many types of environments and difficulty tasks
the robot must be special
To achieve goals with greater efficiency.
Reconfigurable robots can achieve these goals

38. Locomotion modes Each mode consists of
characteristics for the environment type
speed
turning-ability
energy-efficiency
recoverability from failures

39. The 6M-loop mode
6 M-modules are in a ring configuration of hexagon shape
Advantage:
Energy efficient and allows high speeds
Disadvantage:
Tolerance to environment obstacles is limited by the size of the wheel
The robot cannot stand up once it falls down

40. The 6M-loop mode
Shapes alter between a regular hexagon and a deformed hexagon that tends to fall forward.
Starting from the regular hexagon, the movement is controlled by the deformation of the shape to change the centre of gravity of the traveller.
2 commands governing the shape transformation:
One is to retain the regular hexagon shape.
One is to let the rolling traveller to “squeeze” itself to a deformed hexagon.
Commands are selected using gravity sensors

41. The 6M-loop mode

42. The 10C-Loop Mode Uses all CONRO-like modules
each module can control its pitch and yaw movement
Flexible and can run, turn, and recover from falling down
Can deal with environments where obstacles do not exceed in size the height of the robot configuration

43. The 10C-Loop Mode Achieves the rolling track locomotion
At a fixed time interval (OR when all modules have bended forward to the desired angle)
each module begins to bend forward again to reach the angle that is equal to the current angle of the module that is in front of it.
When this process repeats, the rolling track will move forward in a straight path.

44. The 10C-Loop Mode Recovery from fall down

45. The 9M-walker mode
H-Walker is a 4-legged walker using 2 DOF on each module
3 possible local topologies:
Torso, upper leg, and lower leg

46. The 9M-walker mode
Distributed locomotion control was achieved using the digital hormone method [Shen 2002]
4 hormones are used to control each leg
Torso sends the hormone messages to the legs and synchronizes their coordinated actions

47. The 9M-walker mode H-walker mode has symmetric design
Prevents it from falling into any unrecoverable position
Its topology is in the shape of an 'H'
Can walk forwards and backwards using the same strategy

48. The 9M-walker mode
Fall down: It is easy to achieve the relaxed position in which the legs are straightened out to the sides in a double-caterpillar shape.
It stands up using the following steps

49. The 6M4C-training-wheel mode
Modified version of 6M
Added 4 extra legs as “training wheels” to 6M-loop
It can run fast, and can turn and recover from falling

50. The 6M4C-training-wheel mode Recovering from falling
Straightens all the “leg” modules and collapses the hexagon to a flat loop
The hexagon plane can then be made vertical and the flat loop will change back to its hexagon shape and continue to roll

51. The 2M4C-loop mode It uses 6 modules for the loop : MCCMCC
It alternates the types of module to enable the loop to turn and recover from falling

52. The 2M4C-loop mode Recovery from fall down
The loop straightens itself by bending the 2 Mmodules into 180 degrees
Resets the shape of all 4 C-modules
The C-modules then change their yaw servos so that the robot is rising up yet unbalanced.
The unusual movements of the C-modules will cause the robot to fall sideways
The loop will then straighten up again
Goes back to its original hexagon shape

53. The 2M4C-loop mode Recovery from fall down

54. The 8M-climbing mode
8 M-shape Superbot modules forming a rolling track that is only 1.5-module in height
The advantage of this configuration is to make use its low height property to stabilize it on the slope
The mode climbs up the slope slowly by moving module by module

55. The 8M-climbing mode

56. Conclusion
Presents the concept of multimode locomotion for the Superbot robot and a list of locomotion modes
The effectiveness of these modes are demonstrated by the Superbot modules and configurations in simulation
Future work: the process of how to reconfigure the robot from one mode to another through self-reconfiguration

57. References
[Shen 2002] W.-M. Shen, B. Salemi, and P. Will, Hormone-Inspired Adaptive Communication and Distributed Control for CONRO Self-Reconfigurable Robots, IEEE Transactions on Robotics and Automation, 18(5), October, 2002.
[Salemi 2006] Behnam Salami, Mark Moll, and Wei-Min Shen. SUPERBOT: A Deployable, Multi-Functional, and Modular Self-Reconfigurable Robotic System. In Proc. 2006 IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, Beijing, China, October 2006.
[Moll 2006] Mark Moll, Peter Will, Maks Krivokon, and Wei-Min Shen, Distributed Control of the Center of Mass of a Modular Robot,In Proc. 2006 IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, Beijing, China, October 2006.
[Shen 2006] Wei-Min Shen, Maks Krivokon, Harris Chiu, Jacob Everist, Michael Rubenstein, and Jagadesh Venkatesh, Multimode Locomotion via SuperBot Robots, In Proc. 2006 IEEE Intl. Conf. on Robotics and Automation, pp. 2552–2557, Orlando, FL, 2006.

58. Questions